The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through external data sources.
Ingestion points: Merchant names and transaction descriptions retrieved from financial statements via the transaction_search tool described in SKILL.md.
Boundary markers: The instructions in SKILL.md lack specific delimiters or guardrail instructions to ensure the agent ignores malicious commands potentially embedded within transaction strings.
Capability inventory: The skill uses categorize and export_transactions tools mentioned in SKILL.md to map and output data, which could be influenced by injected content.
Sanitization: No string validation or sanitization steps are defined in the skill logic for external inputs.

tax-prep